Introduction

According to WHO, there are 39 million people living with HIV globally in 20221. While once thought of as a deadly disease, HIV is now considered a manageable disease with the advent of highly-active antiretroviral therapy (HAART) in 1995. Now, a growing number of people are aging with HIV. As people age, chronic diseases such as type 2 diabetes mellitus (DM), hypertension (HTN), and other cardio-metabolic diseases may develop2, similar to the general population. People living with HIV are also uniquely susceptible to disease-related vulnerabilities, such as inflammation caused by carrying HIV infection3, impaired immune system4, or toxicities due to long-term antiviral medication use4.

Metabolic syndrome (MetS) is a disorder consisting of obesity, dyslipidemia, hypertension, and insulin resistance. HIV infection has been found to be associated with increased risk of cardiometabolic syndrome in the past3. Recent literature found that age5,6,7,8,9,10,11, body mass index5,6, gender6,8,9, socioeconomic status (as evidenced by education7, wealth8,9, occupation7, residence9), physical activity7,8, use of protease inhibitors2,12,13,14,15,16,17, and CD4 level10,11,17 were risk factors for metabolic syndrome among PLWH. Inflammation (as evidenced by hs-CRP level) was occasionally found to be related to metabolic syndrome18, and smoking may play a role19,20. Most of these studies were done prior to the widespread use of single tablet integrase inhibitors (II) as first line therapy. Besides, there is a paucity of studies to study metabolic syndrome among people living with HIV in Taiwan. In this study, we attempt to delineate the associated factors of metabolic syndrome in a predominant integrase inhibitor era in a Taiwanese population.

Methods

Patient data collection

This is a cross-sectional study involving all people living with HIV (PLWH) aged ≥ 18 years who visited MacKay Memorial Hospital between September 7, 2022 to January 31, 2023. Data were collected in this time interval as most of our patients came back for follow-up every 3 months. For any patients who visited more than one time, only data from the first visit was included. Data collected included smoking, sex, type of HAART used, anti-hepatitis B surface antibody (Anti-HBs), anti-hepatitis B core antigen (Anti-HBc), anti-hepatitis B surface antigen (HBsAg), anti-hepatitis A antibody (Anti-HAV), anti-hepatitis C antibody (HCV), patient’s viral load (VL), rapid plasma reagin (RPR), age, CD4 count, fasting glucose, serum aspartate transaminase (AST), serum alanine transaminase (ALT), serum total cholesterol, triglycerides (TG), creatinine, low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), glycosylated hemoglobin (HbA1c), presence of proteinuria on urinalysis, urine protein to creatinine ratio (UPCR), years since diagnosis, and whether they are on treatment for diabetes mellitus (DM), hyperlipidemia, or hypertension (HTN). Patients with no information on body weight, blood pressure, smoking status, either fasting glucose or HbA1c, or lipid profile were excluded from the study.

We also calculated Homeostatic Model Assessment for Insulin Resistance (HOMA-IR) score21, atherosclerotic cardiovascular disease (ASCVD) risk score22, and Framingham risk score23 in these patients.

For this study, we used a modified National Cholesterol Education Program ATP III definition of metabolic syndrome to define metabolic syndrome. As waist circumference were not routinely collected during office visits, we used body mass index (BMI) as a surrogate for waist circumference, similar to another study in the past12. Metabolic syndrome was defined as the presence of three or more of the following five manifestations: high serum triglycerides, as defined by a triglyceride level of greater or equal to 150 mg/dL; low HDL-C, as defined by a serum HDL-C < 40 mg/dL in man or < 50 mg/dL in woman; hypertension, as evidenced by a clinical diagnosis or in-office blood pressure of a systolic blood pressure ≥ 140 mmHg, or diastolic pressure of ≥ 90 mmHg; obesity, as defined by BMI ≥ 30; and diabetes mellitus, as evidenced by a clinical diagnosis, or laboratory values showing fasting glucose ≥ 126 mg/dL, or HbA1c > 6.5%. Individuals would also be classified as having hypertriglyceridemia, HDL-C hypocholesterolemia, hypertension, or diabetes mellitus if they were receiving specific treatments for these conditions. This study complies with the Declaration of Helsinki and was approved by the MacKay Memorial Institutional Review Board, which waived the need to obtain informed consent from patients. The protocol number from the MacKay Memorial Hospital Institutional Review Board was 23MMHIS280e.

Analysis

Continuous variable of age was grouped into five groups (age 18–30, 31–40, 41–50, 51–60, and 61+). CD4 count was grouped into five groups (CD4 1–50, 51–100, 101–200, 201–400, and more than 400 cell/µL). Serum HIV viral load was grouped into ≤ 200 and > 200 copies/mL. Smoking was a categorical variable with either current smoker, never smoked, or past smoker. Syphilis was categorized as a binary variable of serum RPR positivity or negativity. HAART use was divided into three categories based on the medication’s third agent. The three categories were: (1) regimens that used integrase strand inhibitors (II) (including ABC/3TC/DTG, 3TC/DTG, DTG/RPV, BIC/FTC/TAF, EVG/COBI/FTC/TAF), (2) regimens that included protease inhibitors (PI) (DRV/COBI), and (3) regimens that included non-nucleoside reverse transcriptase inhibitors (NNRTI) (including EFV/TDF/FTC, TDF/3TC/DOR, RPV/TAF/FTC).

Cross tabulation, Chi-squared tests and univariate logistic regression was used to investigate the relationship between categorical variables and metabolic syndrome. Student’s t-test was used for continuous variables. For the factors that were significant in univariate analysis that were considered a possible risk factor, Chi-squared tests were performed to see whether the possible risk factor was associated with any of the independent variables. Those variables that showed evidence of association with both outcome (metabolic syndrome) and the exposure factor were considered to be potential confounders if they were not on the causal pathway. All variables were assessed as potential confounders or effect modifiers by using the Test for Homogeneity. Finally, a final multivariate logistic regression was performed with variables which confounded or modified the association between possible associated factors and metabolic syndrome. All tests were 2-tailed, and p-values of less than 0.05 was considered significant. The statistical analyses were performed either using SPSS 21.0 (SPSS Inc., Chicago, IL, USA) or STATA 16 (StataCorp LLC., College Station, TX, USA).

Results

A total of 1130 PLWH were followed in our hospital. After excluding 321 people who lacked data in either height, body weight, smoking, blood pressure, fasting glucose, HbA1c or lipid profile, a total of 809 were included into the study (Fig. 1). We compared gender and age between people that were included or excluded, and no significant difference was found. (Table 1)

Fig. 1
figure 1

Flow chart of included participants.

Full size image
Table 1 Characteristics of study group and excluded group.
Full size table

In our study, 81 people had been diagnosed with metabolic syndrome, resulting in a prevalence rate of 10.0%. The most common HAART regimen was II-based (81.3%), which included BIC/FTC/TAF, 3TC/DTG, EVG/COBI/FTC/TAF, DTG/RPV, and ABC/3TC/DTG. 96.0% of our patients had a viral load below 200 copies/ml, and 75.0% had a CD4 count above 400 cells/µl. There were 11 treatment-naïve PLWH and two of them were diagnosed with metabolic syndrome. (Table 2)

Table 2 Epidemiological and clinical characteristics of PLWH and odds ratio of metabolic syndrome.
Full size table

Comparing the characteristics between the group with and without metabolic syndrome, there were no significant differences in the distribution of sex (p = 0.711), smoking (p = 0.737), HAART regimen (p = 0.240), VL (p = 0.686), CD4 level (p = 0.620), treatment experience (p = 0.34), RPR (p = 0.84), Anti-HAV (p > 0.999), Anti-HBs (p = 0.628), HBsAg (p = 0.168), Anti-HCV (p = 0.509), or duration of treatment (p = 0.075). MetS group had higher prevalence of Anti-HBc positivity (p = 0.05), HTN (p < 0.001, DM (p < 0.001), hypertriglyceridemia (p < 0.001), HDL-C hypocholesterolemia (p < 0.001), and obesity (p < 0.001). Those in MetS group were also older (p < 0.001). In biochemical data, level of AST (p = 0.139), total cholesterol (p = 0.992), creatinine (p = 0.930), LDL-C (p = 0.154), UPCR (p = 0.318), and hs-CRP (p = 0.115) were not significantly different. (Table 2) Those in MetS group had higher fasting glucose (mg/dL) (p < 0.001), HbA1c (%) (p < 0.001), but lower HDL-C (mg/dL) (p < 0.001). Those in MetS group also had higher scores in HOMA-IR (p < 0.001), ASCVD risk (p < 0.001), and Framingham risk (p < 0.001). (Table 2)

Initially, we divided the HAART regimen into three groups according to third agents and no statistical significance was noted (p = 0.240). (Table 2) To further explore this relationship, we regrouped HAART regimen into six smaller groups, as detailed in Table 3, and none of them showed statistical significance compared with BIC/FTC/TAF (p = 0.403, 0.195, 0.286, 0.219, 0.541, respectively).

Table 3 Odds ratio of detailed HAART regimen.
Full size table

In the group with metabolic syndromes, the most prevalent component among the five diagnostic criteria of metabolic syndrome was HDL-C hypocholesterolemia, which was found in 69 (85.2%) people of the group. The next common condition was hypertension (84.0%), followed by hypertriglyceridemia (81.5%). Less than half of this group had obesity or DM (42.0% and 38.3% respectively). (Table 2) Among the components of metabolic syndrome, obesity had the strongest association with metabolic syndrome (OR: 32.19, CI: 16.58–62.50), followed by hypertension (OR: 19.33, CI: 10.41–35.92), HDL-C hypocholesterolemia (OR: 16.40, CI 8.69–30.99) and hypertriglyceridemia (OR: 15.13, CI 8.41–27.22). Although DM was the most weakly associated with metabolic syndrome, having DM was still linked with an over tenfold risk of developing metabolic syndrome (OR = 14.94; CI: 8.35–26.74).

Among factors that showed significant association with metabolic syndrome, some were related to the diagnostic criteria of metabolic syndrome and thus association was expected. Age was not one of the diagnostic criteria, but it showed strong association with metabolic syndrome.

After performing tests of homogeneity, it was found that fitting the other independent variables (sex, smoking, HAART group, viral load group, CD4 level, RPR, Anti-HAV, Anti-HBs, HBsAg, Anti-HCV, duration of treatment) into the multivariate logistic model did not change the odds ratios significantly or failed due to issues of collinearity. Current smoker (p = 0.032), age over 61 years old (p = 0.013) and CD4 lower than 200 cells/µl (p = 0.008) showed significance. (Table 4)

Table 4 Odds ratio of multivariate logistic regression.
Full size table

Discussion

The global prevalence of metabolic syndrome in PLWH is approximately 16.7%.19 Many studies have showed that the prevalence rates of metabolic syndrome in PLWH varied from 19.2–40.1%.5,8,11,12,18,24,25,26,27 These variations could be attributed to differences in regional diets, genetic predispositions, socioeconomic factors, and local healthcare practices.

In our study, the prevalence of metabolic syndrome in PLWH was 10.0%, which was lower than the global average and other documented prevalences. In similar studies in Asia, the prevalence rate was 23.6% in Singapore12 and 40.1% in India5. A previous study in Taiwan in 2012 reported a prevalence of metabolic syndrome in PLWH of 26.2%.27 In the study in Singapore by Ang et al., researchers analyzed 2231 treatment-experienced PLWH. Most of them were men (93.9%). All of them had been exposed to NRTI as the first line of treatment, 93.9% to NNRTI, 28.6% to protease inhibitors and 12.8% to integrase strand transfer inhibitors12. In the study by Mally et et al., 182 PLWH were included, both treatment-experienced and treatment-naïve. Most of them were using NRTI and NNRTI5. In the study by Wu et al., 877 PLWH were included. 81.7% of them were using NRTI as HAART but no one used II for treatment27. We suspected that the marked discrepancy between our number and those documented previously could be, at least in part, attributed to the evolution in the choice of HAART regimens. In the past, we often used protease inhibitors to suppress HIV. PIs had been reported to influence lipid and glucose metabolism, which may lead to metabolic syndrome15,16,28. On the other hand, use of integrase inhibitors was associated with a lower risk of metabolic syndrome2. As the Guidelines for diagnosis and treatment of HIV/AIDS, published by Taiwan AIDS society, suggested the usage of II-based regimen as first-line HAART since 201629, most of our patients were using II-based regimen (81.3%) instead of PI-based regimen (0.9%). The fact that our prevalence rate was lower than previously reported could be the result of a fall in the prevalence of metabolic syndrome after the adoption of II-based regimen, despite the weight gain implications of these medications30,31. This finding is comparable with previous finding from Taramasso et al.32.

In previous research, the duration of an HIV diagnosis, the severity of the infection, and the length of HAART usage have been identified as risk factors for metabolic syndrome2,17,18,27. Associations have been drawn between an increased risk of metabolic syndrome and virological failure (viral load > 1000 copies/ml)18, low CD4 counts11, and long-term HAART use27. Our study didn’t find a strong correlation between CD4 levels or viral load and metabolic syndrome. Additionally, we observed no correlation between the duration of HIV diagnosis (as a proxy for HAART duration) and metabolic syndrome. Using protease inhibitors as part of HAART has been reported as a risk factor of metabolic syndrome in previous study2,4,15. In our study, the use of protease inhibitors was not significantly associated with metabolic syndrome (OR: 3.59, CI: 0.68–18.86). The wide confidence interval may be attributed to a small subgroup sample size (n = 7). Thus, although we did not find statistically significant association between protease inhibitors and metabolic syndrome, our study is still in line with previous research.

On reviewing the medication history of our patients, we found that the medication use among our patients conformed to the recommended treatment available during different time periods. Around 2005, patients were on an NRTI plus a protease inhibitor such as nelfinavir. Nelfinavir was then changed to regimen containing atazanavir or lopinavir/ritonavir around 2007. Some patients changed to raltegravir around 2015, then some changed to rilpivirine-based regimen while most others changed to dolutegravir-based regimen in 2017 or 2018. When bictegravir became available in 2019 this became a popular choice among many patients. As our patients were quite heterogeneous in switching regimens, it was not possible to make a true comparison to determine the effects of switching individual regimens. Hence our present study shows only a snapshot of the prevalence of metabolic syndrome at a point in time, and hope this may provide a basis from which to base future comparisons.

In our study, age remains one of the factors associated with metabolic syndrome33,34,35. Only age showed significant difference in both univariate analysis and multivariate logistic regression analysis. It had a strong positive correlation with HTN (p < 0.001), DM (p < 0.001), hypertriglyceridemia (p < 0.001), and HDL-C hypocholesterolemia (0.003). (Table 5) Consequently, the likelihood of developing metabolic syndrome increases as individuals age.

Table 5 Prevalence of HTN, DM, hypertriglyceridemia, HDL-C hypocholesterolemia, obesity and metabolic syndrome in different age groups.
Full size table

Limitation

Our study had a few limitations. First, our study is a single-institution study. The results may not accurately represent the prevalence of metabolic syndrome among the PLWH in Taiwan. Second, as this is a cross sectional analysis, causality could not be inferred. Third, we cannot fully evaluate the implication of prior HAART use, as our patients often relocate or change institutions of care. However, prior to the adoption of IIs as first-line treatment, most of our patients were on an NNRTI-based regimen as first-line or on PIs if they had virological resistance. It can be seen from the study that treatment duration was not associated with metabolic syndrome, hence the use of prior types of HAART regimen likely did not affect the outcome of the study. Finally, there could be residual confounding that are not accounted for within the scope of our study.

Conclusion

In conclusion, our findings indicate a lower prevalence rate compared to global studies, which may be related to broad adoption of II-based HAART. PI-based HAART appeared to be a risk factor, but it did not reach statistical significance in this study.

Age was strongly associated with metabolic syndrome. It was also associated with other factors (HTN, DM, hypertriglyceridemia, HDL-C hypocholesterolemia) in our study. The insights from our research can guide both clinicians and patients to focus on managing traditional risk factors, including hypertension, diabetes mellitus, dyslipidemia, and obesity, to prevent metabolic syndrome and its associated complications.